Re: Weird bitset shift offset in zend_alloc

Hi! > This is intended x86 specific optimization. The behavior of SHR is > documented. > > See https://c9x.me/x86/html/file_module_x86_id_285.html > > "The count is masked to 5 bits, which limits the count range to 0 to 31" > on 32-bit systems. I see. But I wonder whether the compiler will always compile it as SHR, is it guarnateed? Looking at optimization behavior for clang, for me ZEND_BIT_TEST( bits, 65 ) - where bits is an array of 0xFF - produces 1 under -O0 but 0 under -O1 (I guess it figures out it's a constant and optimizes the shift away). So I wonder whether this construct isn't dangerous... -- Stas Malyshev smalyshev@wikimedia.org